[C++ PATCH]:Fix 13387 breakage

[Richard Sandiford]

Thanks for the quick reponse!  Your patch certainly fixed the irix problems.

Nathan Sidwell <nathan@codesourcery.com> writes:
> This patch fixes the aliasing breakage that Richard Sandiford
> pointed out from my 13387 patch. I have examined the mips assembly this
> produces and it seems good, but I am waiting for Richard to tell me that
> the test case confirms it, before checking that in[1].

Nearly.  Although the assembly code for the test was wrong before the patch,
it still passed execution, presumably because the uninitialised f1(x).index
happened to be 0.  The version below failed before your patch but passes
after it.

Also, not all mips targets support -mabi=64, and those that do need
to test it as a separate multilib.  It might be better to remove
the mips-specific option and make it a general test for all targets.

Does this version look OK?

Richard

-------------- next part --------------
A non-text attachment was scrubbed...
Name: test.cpp
Type: text/x-c++src
Size: 649 bytes
Desc: not available
URL: <http://gcc.gnu.org/pipermail/gcc-patches/attachments/20040104/eb661e4a/attachment.bin>